Terjman-Ultra-v2.0
| Property | Value |
|---|---|
| Parameter Count | 1.3B |
| Base Model | facebook/nllb-200-1.3B |
| License | CC BY-NC |
| BLEU Score | 25.00 |
| Training Precision | Mixed FP16 |
What is Terjman-Ultra-v2.0?
Terjman-Ultra-v2.0 is a state-of-the-art translation model specifically designed for English to Moroccan Darija translation. Built upon the facebook/nllb-200-1.3B architecture, this model represents a significant improvement over its predecessor, achieving impressive BLEU scores of 25.00 and competing with much larger proprietary models.
Implementation Details
The model leverages a Transformer-based sequence-to-sequence architecture, trained using carefully curated parallel corpora. It implements mixed FP16 precision for efficient inference and was trained with sophisticated hyperparameters including a learning rate of 0.0005, batch size of 128, and AdamW optimizer.
- Fine-tuned specifically for English->Moroccan Darija translation
- Achieves near-proprietary model performance levels
- Optimized for cultural accuracy and natural language flow
- Compatible with 🤗 Transformers library
Core Capabilities
- High-quality translation with BLEU score of 25.00
- Efficient inference with FP16 precision
- Handles complex cultural nuances
- Easy deployment options including API integration
Frequently Asked Questions
Q: What makes this model unique?
Terjman-Ultra-v2.0 stands out for its specialized focus on English-Moroccan Darija translation, achieving performance metrics that approach proprietary models while maintaining open-source accessibility. Its careful optimization for cultural accuracy sets it apart from general-purpose translation models.
Q: What are the recommended use cases?
The model is ideal for research and personal projects requiring high-quality English to Moroccan Darija translation. It's particularly well-suited for applications needing cultural sensitivity in translation. Note that commercial use requires special permission due to the CC BY-NC license.
